The best value for ADJUSTED mcfadden

[Gns Met]

I know that the best values for r2 mcadden are between 0.2 and 0.6, but what about the adjusted macfadden R2?

[Nick Cox]

Hard to know what to say except the higher the better (without other information).

If I am predicting frost or not from temperature with logit, I expect high values; otherwise I messed up.

If I am doing social research high predictability means that I have chosen an utterly banal research question.

Please note our longstanding request to use full real names:

https://www.statalist.org/forums/help#realnames

https://www.statalist.org/forums/help#adviceextras #3

[Nick Cox]

Cross-posted in slightly different form at https://stats.stackexchange.com/ques...en-s-r-squared

Please note our explicit policy on cross-posting: you are asked to tell us about it.

[Gns Met]

Cross-posted in slightly different form at https://stats.stackexchange.com/ques...en-s-r-squared

Please note our explicit policy on cross-posting: you are asked to tell us about it.

Thank you very much Nick, in my case i got 0.204 on McFadden's R2 and 0.177 on McFadden's Adj R2, Do you think its a bad model?

[Mike Lacy]

Either McFadden's R2 is arguably a bad measure from the perspective of reproducing the R2 for an underlying continuous variable truncated into a binary. See:
Demaris, A. 2002. "Explained Variance in Logistic Regression: A Monte Carlo Study of Proposed Measures. Sociological Methods and Research. 31: 27-74.

If I recall correctly, McKelvey and Zavoina's measure, and the Cragg-Uhler measure are among the best.
And I like Nick's points about high and low predictability in social research.

[Nick Cox]

Turn and turn about, I agree with Mike.

On #4 what kind of results do similar exercises in your unstated field return? Sometimes a poor model result indicates an incompetent researcher, sometimes just a refractory problem. Although these measures are figures of merit to say how well you did, they require context for interpretation.